Application Of Memristor Neural Network Based On PPA In Robot Path Planning

Path planning is a key technology in mobile robot control.How to efficiently complete the path planning in static and dynamic environment is always a difficult and hot topic.Based on the characteristics of "integrated storage" of memristor,this thesis proposed the combination of Physarum polycephalus algorithm and memristor array,which effectively solved the problem of path planning in static environment.Combining fuzzy control algorithm with memristor neural network,the path planning problem of mobile robot in dynamic environment is solved effectively.The main work is as follows:In Chapter one,the research background and significance of mobile robot path planning are presented.Firstly,the common path planning algorithms in recent years are introduced from two aspects: global path planning and local path planning,and their advantages and limitations are analyzed.As the complexity of the mobile robot’s working environment increases,the data needed to be processed by the algorithm becomes more and more large.The memristor’s "integrated storage" feature can solve this problem,thus speeding up the convergence speed of the algorithm.In order to further understand the memristor,this chapter then from the proposed memristor,the main three application fields and the main five kinds of mathematical models for a more comprehensive analysis.In Chapter two,based on the analysis of the main research and the basic algorithm to be improved,the basic thesis,development history and research status of traditional Physarum polycephalus algorithm,fuzzy control,neural network and memristor are introduced.The parameters of the memristor model used in this paper are set and simulated.In Chapter three,the path planning of mobile robot in static environment is studied.The traditional Physarum polycephalus algorithm uses the form of solving equations to calculate the pressure,which requires multiple matrix transformations in each iteration,resulting in too much computation.In this thesis,the method of independently calculating the pressure of Physarum polycephalus algorithm makes the iterative calculation of pressure only rely on local information,instead of solving a large number of equations.Thus,the computational complexity of Physarum polycephalus algorithm is reduced effectively.Secondly,based on Kirchhoff’s law,the algorithm is updated and iterated in the memristor array using read and write operations,and then the convergence of the algorithm is judged by the memristor resistance value and the current flowing through each memristor.The simulation results show that the proposed algorithm combining Physarum polycephalus algorithm with memristor array converges faster than the traditional algorithm,and provides a theoretical basis for parallel calculation of memristor array.In Chapter four,aiming at the problem of collision avoidance of mobile robots in dynamic environment,this thesis uses memristor fuzzy neural network to achieve real-time obstacle avoidance.The speed of the dynamic obstacle and its distance from the mobile robot are the input of fuzzy control,and the rotation Angle of the mobile robot is the output.The parts of fuzzy inference,fuzzy inference and defuzzification are represented by memristor array structure.As the neurons of the memristor fuzzy neural network,dynamic obstacle avoidance is carried out in the memristor fuzzy neural network.In order to verify the feasibility of the algorithm,four dynamic obstacles are set up in the experimental environment.The experimental results show that the algorithm can make the mobile robot safely avoid the dynamic obstacles and successfully reach the target point.